Fig. 7

The 2000 wave 1 run outputs f _i (x) for all 18 outputs considered (see Eq. (22)) are shown as the purple lines, with the observed data z _i ±3σ _i given as the black error bars, and the best run previously found by [30] shown as the light blue line. The horizontal black line at zero represents no trend. As these runs were generated from a space filling maximin Latin hypercube design, they can give substantial insight into the broad behaviour of the model over the initial search region \(\mathcal {X}_{1}\). We can see that some outputs are seemingly constrained to give only positive (e.g. Auxin_fa) or negative (e.g. Auxin_mu) trends, and that many of the runs are far from the target ranges (as the y-axis is on a log scale). We also find that no individual wave 1 run passes through every one of the target intervals